To identify differentially expressed proteins amongst cDC subsets we used a mass spectrometry based proteomics approach to quantitatively compare FACS purified cDC subsets of the spleen of uninfected steady-state mice. cDCs were subsequently purified from Collagenase and DNAse digested spleens by density centrifugation and magnetic depletion of non-cDC resulting in a 70-80% enriched cDC cell population, which was segregated into subsets according to the expression of CD8 and CD4 surface molecules using flow cytometry (Figure 2-1A). cDCs preparations from pooled spleens of 32 mice consistently yielded more than 2.5x106 cDCs per subset with purity higher than 95%, as estimated by FACS-reanalyses (Figure 2-1B).
Sorted CD4+, CD8α+ and DN cDCs, resulting in 20 to 25 g total protein each, were separated by 1D sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and after tryptic in-gel digestion analyzed by LC-MS/MS (LTQ-Orbitrap). We repeated this experiment twice with different pools of spleens resulting in biological triplicates.
Dendritic cells differ from viral recognition 32
LC-MS/MS data from all 126 gel slices of the three independent large scale experiments were combined and analyzed by MaxQuant (Cox and Mann, 2008).
Figure 2-1 Experimental strategy for in vivo cDC subset proteomics (A) and FACS purification (B)
We incorporated new algorithms for label-free quantification into MaxQuant, which allows us to quantify peptides across individual LC-MS/MS runs (Cox et al., submitted and attached). Analysis of the entire dataset using uniform statistical criteria identified 1,283,676 peptides (99,228 non-redundant sequences), corresponding to 5359, 5642 and 4830 proteins in CD4+, CD8α+ and DN cDCs with 99% certainty, respectively.
Summed peptide intensity is a good proxy for absolute protein abundance (de Godoy et al., 2008b) and by this measure we covered the cDC proteomes across more than four orders of magnitude. The cDC proteome of 5780 proteins shows no bias against low level regulatory proteins such as kinases (209 identified or 3.6% of our proteome vs. 4.2% in the genome). Relative label-free quantitation was highly reproducible between biological replicates and correlation between normalized protein intensities was between 0.84 and 0.96 (Figure 2-2). Hence, label free quantitative proteomics provides an accurate means of comparing differences in protein expression between cDC subsets in vivo for thousands of proteins.
Figure 2-2 Reproducibility of biological and technical replicate measurements of (A) CD4+ cDCs, (B) CD8α+ cDCs and (C) DN cDCs. Correlation is determined by Pearson coefficient
Expression levels of most proteins were similar and only a relatively small number showed statistically significant differences between any two subsets (p ≤ 0.01) (Figure 2-3). Overall,
Dendritic cells differ from viral recognition 34
differences in protein expression patterns showed that CD4+ and DN cDCs were more closely related to each other than to CD8 + cDCs (Figure 2-3) as already suggested by microarray analysis (Edwards et al., 2003a).
Figure 2-3 Volcano plot of protein expression differences between cDC subsets as a function of statistical significance. Proteins with no detectable signal in one of the subsets were assigned a ratio of infinity. (A) CD4+ vs. CD8α+ cDCs, (B) CD4+ vs. DN cDCs and (C) DN vs. CD8α+ cDCs
We then analyzed the correlation between label-free quantitation results and known marker proteins of cDC subsets. All subsets expressed CD11c, the pan-marker of DCs. Markers for other lymphocyte lineages like CD19 for B lymphocytes, CD3 subunits for T lymphocytes or Nk1.1 for NK cells were not detected, indicating that sorted cDC populations were indeed pure and did not contain detectable contaminations from other cell types. The surface
markers CD4 and CD8α, which were used for FACS purification were expressed by the respective population and correctly assigned by our bioinformatic algorithms. Importantly, this was also true for other known surface markers such as DEC-205 and CD36 for CD8α+ DC, and CD11b, 33D1 and SIRPα for CD4+ and DN cDCs (Table 2-1). Correct assignment of these key surface markers by hypothesis free quantitative label free proteomics demonstrates the robustness of our approach. TLR3, an endosomal TLR, provided a further positive control (Edwards et al., 2003b) (Table S2-1 online).
Table 2-1 Numbers are the median of summed peptide intensities in ion counts per second over three biological replicates. CD4+ cDCs CD8 + cDCs DN cDCs CD11c 78,828,000 58,823,000 55,484,000 CD8α 0 3,876,600 0 CD205 381,840 6,461,300 443,580 CD36 92,491 3,421,300 445,720 CD4 1,096,000 0 0 Sirpα 6,585,700 139,680 4,044,100 33D1 950,990 0 694,500 CD11b 17,304,000 1,211,600 16,806,000
Subset restricted transcription factors distributed as expected; for example interferon regulatory factor 8 (IRF-8) was specific to CD8 + compared to CD4+ DCs, whereas IRF-4 and the NF- B subunit RelB were much more abundant in CD4+ DCs (Table S2-1 online). Interestingly, the class I NF- B family members Nf B1 and Nf B2 have a similar expression pattern to RelB while RelA and c-Rel have no subtype specificity, suggesting a functional specialization of the NF B pathway in cDC subtypes, previously only suggested by differential effects on DC subsets in mutant mice lacking members of the NF- B family (O'Keeffe et al., 2005b; Wu et al., 1998) .
CD97, a member of the EGF-TM7 family of adhesion receptors implicated in leukocyte homing, was four-fold more abundant in CD8α+ compared to CD4+ cDCs. Flow cytometry with antibodies against CD97 independently confirmed our proteomic finding (Figure 2-4). In contrast, previous microarray data did not classify CD97 as a surface molecule with significant higher expression on CD8α+ cDC (Dudziak et al., 2007; Edwards et al., 2003a).
Dendritic cells differ from viral recognition 36
Proteome and transcriptome measurements do not necessarily agree, either due to technical factors or due to additional regulation at the protein level (Bonaldi et al., 2008).